[Pw_forum] What happens at REALLY large ectuwfc?
>start with 'random' or 'atomic+random' initial wavefunctions >(startingwfc='...'). I didn't find anything wrong with very >high cutoffs. Occasionally you can end up in the wrong ground >state, though, especially in highly symmetric cases like the >Si example you sent. Hi Paolo. I tried startingwfc='random' and the eigenvalues are now essentially the same as the 200 Ry calculation. k = 0. 0. 0. (** PWs) bands (ev): -5.1746 7.0369 7.0369 7.0369 !total energy = -14.59208472 Ry Harris-Foulkes estimate = -14.59208472 Ry estimated scf accuracy<3.8E-11 Ry The total energy is the sum of the following terms: one-electron contribution = 5.58227211 Ry hartree contribution = 1.67255887 Ry xc contribution =-5.04795092 Ry ewald contribution= -16.79896478 Ry Fock energy 1 = 0. Ry Fock energy 2 = 0. Ry Half Fock energy 2= 0. Ry Thanks, Brad On Sun, Jan 24, 2010 at 2:38 PM, Brad Malone wrote: > >are you using the latest cvs version? apparently there is a problem > >with the new symmetrization algorithm that will be fixed ASAP. > > The results I posted are from espresso-4.0.5, although I originally saw > this problem with espresso-4.1.1 in a different system (AlAs on a 2x2x2 > shifted grid). > > As for what Lorenzo said, it makes sense with what I'm seeing. The energy > breakdowns for the 200 Ry and the 2000 Ry cases are shown below: > > For 200 Ry: > > !total energy = -14.59208467 Ry > Harris-Foulkes estimate = -14.59208467 Ry > estimated scf accuracy<3.9E-11 Ry > > The total energy is the sum of the following terms: > > one-electron contribution = 5.58227319 Ry > hartree contribution = 1.67255719 Ry > xc contribution =-5.04795028 Ry > ewald contribution= -16.79896478 Ry > Fock energy 1 = 0. Ry > Fock energy 2 = 0. Ry > Half Fock energy 2= 0. Ry > - > For 2000 Ry: > - > !total energy = -14.11008918 Ry > Harris-Foulkes estimate = -14.11008918 Ry > estimated scf accuracy<1.0E-11 Ry > > The total energy is the sum of the following terms: > > one-electron contribution = 6.07256114 Ry > hartree contribution = 1.58024935 Ry > xc contribution =-4.96393489 Ry > ewald contribution= -16.79896478 Ry > Fock energy 1 = 0. Ry > Fock energy 2 = 0. Ry > Half Fock energy 2= 0. Ry > --- > > So you can see the one-electron contribution going up and the hartree > contribution going down as Lorenzo as argued. > > Brad > -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20100125/4075c9ae/attachment.htm
[Pw_forum] What happens at REALLY large ectuwfc?
I would be surprised if roundoff errors in FFT should be blamed for the reported behavior SB On Jan 25, 2010, at 11:34 AM, O. Baris Malcioglu wrote: > Dear Brad, > > Apart from what is already mentioned, another contribution to error > you are mentioning might be due to computational/numeric issues. > > Run-of-the-mill FFT algorithms has an error upperbound that goes like > O(sqrt(N)). Even in the best case (least performing) scenarios the > error upper bound is O(log(N)). > > These errors are due to numeric precision employed, and mostly bound > by hardware (unless you compromise performance significantly). > > In general, in any calculation that involves computers, in a graph > that shows the "error" as a function of some convergent parameter, I > would expect three regions , first the error diminishes due to > numerical reasons, then there is a flat region, then roundoff errors > begin to dominate, and error picks up again. > > If interested, please have a look at the (rather rusty but still > valid) concept "Machine constant" > > Best, > Baris. > > > > 2010/1/25 Paolo Giannozzi : >> Brad Malone wrote: >>> I originally saw this problem with espresso-4.1.1 in a >>> different system (AlAs on a 2x2x2 shifted grid). >> >> start with 'random' or 'atomic+random' initial wavefunctions >> (startingwfc='...'). I didn't find anything wrong with very >> high cutoffs. Occasionally you can end up in the wrong ground >> state, though, especially in highly symmetric cases like the >> Si example you sent. >> >> "Ghost states" are another story. It may happen that a >> (nonlocal, separable) pseudopotential has a localized >> spurious state that doesn't show up (i.e. it is not >> occupied) at low cutoff, but it is occupied at higher >> cutoff. None of the PP on the QE web site should have >> such states, though. >> >> If you have evidence of further problems, please provide >> a test case so that it can be reproduced. In any case, >> thank you for finding a bug in the cvs version! >> >> P. >> --- >> Paolo Giannozzi, Democritos and University of Udine, Italy >> ___ >> Pw_forum mailing list >> Pw_forum at pwscf.org >> http://www.democritos.it/mailman/listinfo/pw_forum >> > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum --- Stefano Baroni - SISSA & DEMOCRITOS National Simulation Center - Trieste http://stefano.baroni.me [+39] 040 3787 406 (tel) -528 (fax) / stefanobaroni (skype) La morale est une logique de l'action comme la logique est une morale de la pens?e - Jean Piaget Please, if possible, don't send me MS Word or PowerPoint attachments Why? See: http://www.gnu.org/philosophy/no-word-attachments.html -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20100125/a1a0b8ae/attachment.htm
[Pw_forum] What happens at REALLY large ectuwfc?
Dear Brad, Apart from what is already mentioned, another contribution to error you are mentioning might be due to computational/numeric issues. Run-of-the-mill FFT algorithms has an error upperbound that goes like O(sqrt(N)). Even in the best case (least performing) scenarios the error upper bound is O(log(N)). These errors are due to numeric precision employed, and mostly bound by hardware (unless you compromise performance significantly). In general, in any calculation that involves computers, in a graph that shows the "error" as a function of some convergent parameter, I would expect three regions , first the error diminishes due to numerical reasons, then there is a flat region, then roundoff errors begin to dominate, and error picks up again. If interested, please have a look at the (rather rusty but still valid) concept "Machine constant" Best, Baris. 2010/1/25 Paolo Giannozzi : > Brad Malone wrote: >> I originally saw this problem with espresso-4.1.1 in a >> different system (AlAs on a 2x2x2 shifted grid). > > start with 'random' or 'atomic+random' initial wavefunctions > (startingwfc='...'). I didn't find anything wrong with very > high cutoffs. Occasionally you can end up in the wrong ground > state, though, especially in highly symmetric cases like the > Si example you sent. > > "Ghost states" are another story. It may happen that a > (nonlocal, separable) pseudopotential has a localized > spurious state that doesn't show up (i.e. it is not > occupied) at low cutoff, but it is occupied at higher > cutoff. None of the PP on the QE web site should have > such states, though. > > If you have evidence of further problems, please provide > a test case so that it can be reproduced. In any case, > thank you for finding a bug in the cvs version! > > P. > --- > Paolo Giannozzi, Democritos and University of Udine, Italy > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum >
[Pw_forum] What happens at REALLY large ectuwfc?
Brad Malone wrote: > The results I posted are from espresso-4.0.5, although > I originally saw this problem with espresso-4.1.1 in a > different system (AlAs on a 2x2x2 shifted grid). start with 'random' or 'atomic+random' initial wavefunctions (startingwfc='...'). I didn't find anything wrong with very high cutoffs. Occasionally you can end up in the wrong ground state, though, especially in highly symmetric cases like the Si example you sent. "Ghost states" are another story. It may happen that a (nonlocal, separable) pseudopotential has a localized spurious state that doesn't show up (i.e. it is not occupied) at low cutoff, but it is occupied at higher cutoff. None of the PP on the QE web site should have such states, though. If you have evidence of further problems, please provide a test case so that it can be reproduced. In any case, thank you for finding a bug in the cvs version! P. --- Paolo Giannozzi, Democritos and University of Udine, Italy
[Pw_forum] What happens at REALLY large ectuwfc?
Lorenzo: your (ir-)rationale may be more or less justified for separable potentials, but it is certainly wrong for good ol' semilocal ones. SB On Jan 24, 2010, at 9:05 PM, Lorenzo Paulatto wrote: > I'm reposting here a message that I mistakenly sent privately to Brad: > > have a look at the total-energy components: if the one-electron > contributions increases at the expenses of the other terms (mostly > the hartree term) than you have found a ghost in the pseudopotential. > As far as I know, all non-local separable pseudopotentials have ghost > states for large enough values of the curoff. > > I'm not really sure my impression is correct, but I've always had > problems with large enough cutoffs with any pseudopotential I have > tried. My rationale is more or less on this line: the pseudopotential > only have Fourier components up to a certain threshold; plane waves > over that threshold do not feel any direct effect from the ions. As a > consequence this very high frequency plane waves gain nothing by > forming a charge density close to the ions, instead they can just > spread around to minimize the Hartree energy (which decreases) at the > expenses of kinetic energy (which increases). > > cheers > > > -- > Lorenzo Paulatto > SISSA & DEMOCRITOS (Trieste) > phone: +39 040 3787 511 > skype: paulatz > www: http://people.sissa.it/~paulatto/ > > *** save italian brains *** > http://saveitalianbrains.wordpress.com/ > > > > SISSA Webmail https://webmail.sissa.it/ > Powered by Horde http://www.horde.org/ > > > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum --- Stefano Baroni - SISSA & DEMOCRITOS National Simulation Center - Trieste http://stefano.baroni.me [+39] 040 3787 406 (tel) -528 (fax) / stefanobaroni (skype) La morale est une logique de l'action comme la logique est une morale de la pens?e - Jean Piaget Please, if possible, don't send me MS Word or PowerPoint attachments Why? See: http://www.gnu.org/philosophy/no-word-attachments.html -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20100124/d9e92b42/attachment-0001.htm
[Pw_forum] What happens at REALLY large ectuwfc?
I'm reposting here a message that I mistakenly sent privately to Brad: have a look at the total-energy components: if the one-electron contributions increases at the expenses of the other terms (mostly the hartree term) than you have found a ghost in the pseudopotential. As far as I know, all non-local separable pseudopotentials have ghost states for large enough values of the curoff. I'm not really sure my impression is correct, but I've always had problems with large enough cutoffs with any pseudopotential I have tried. My rationale is more or less on this line: the pseudopotential only have Fourier components up to a certain threshold; plane waves over that threshold do not feel any direct effect from the ions. As a consequence this very high frequency plane waves gain nothing by forming a charge density close to the ions, instead they can just spread around to minimize the Hartree energy (which decreases) at the expenses of kinetic energy (which increases). cheers -- Lorenzo Paulatto SISSA & DEMOCRITOS (Trieste) phone: +39 040 3787 511 skype: paulatz www: http://people.sissa.it/~paulatto/ *** save italian brains *** http://saveitalianbrains.wordpress.com/ SISSA Webmail https://webmail.sissa.it/ Powered by Horde http://www.horde.org/
[Pw_forum] What happens at REALLY large ectuwfc?
On Jan 23, 2010, at 5:16 , Brad Malone wrote: > So besides the using of a 2000 Ry cutoff for silicon, what else is > wrong here? are you using the latest cvs version? apparently there is a problem with the new symmetrization algorithm that will be fixed ASAP. By the way, fcc Si with k=0 is an unphysical system: it may easily yield funny results P. --- Paolo Giannozzi, Dept of Physics, University of Udine via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222
[Pw_forum] What happens at REALLY large ectuwfc?
>are you using the latest cvs version? apparently there is a problem >with the new symmetrization algorithm that will be fixed ASAP. The results I posted are from espresso-4.0.5, although I originally saw this problem with espresso-4.1.1 in a different system (AlAs on a 2x2x2 shifted grid). As for what Lorenzo said, it makes sense with what I'm seeing. The energy breakdowns for the 200 Ry and the 2000 Ry cases are shown below: For 200 Ry: !total energy = -14.59208467 Ry Harris-Foulkes estimate = -14.59208467 Ry estimated scf accuracy<3.9E-11 Ry The total energy is the sum of the following terms: one-electron contribution = 5.58227319 Ry hartree contribution = 1.67255719 Ry xc contribution =-5.04795028 Ry ewald contribution= -16.79896478 Ry Fock energy 1 = 0. Ry Fock energy 2 = 0. Ry Half Fock energy 2= 0. Ry - For 2000 Ry: - !total energy = -14.11008918 Ry Harris-Foulkes estimate = -14.11008918 Ry estimated scf accuracy<1.0E-11 Ry The total energy is the sum of the following terms: one-electron contribution = 6.07256114 Ry hartree contribution = 1.58024935 Ry xc contribution =-4.96393489 Ry ewald contribution= -16.79896478 Ry Fock energy 1 = 0. Ry Fock energy 2 = 0. Ry Half Fock energy 2= 0. Ry --- So you can see the one-electron contribution going up and the hartree contribution going down as Lorenzo as argued. Brad -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20100124/059131a8/attachment.htm
[Pw_forum] What happens at REALLY large ectuwfc?
Hi Brad, probably the calculation is not converged. Try a few things: 1) is the total energy at 2000Ry lower than at 200 Ry ? At self-consistency, it should be, even infinitesimally, due to the variaitonal principle. 2) try different recipes: cg vs davidson, and different initalizations (atomic+random, random, etccc). nicola Brad Malone wrote: > I happened to stumble upon something today that I thought was > unusual. The problem is illustrated in a simple example: if you take > the following basic cubic Si input file and, say, the Si.pz-vbc.UPF > pseudopotential from the QE website and do a test for convergence with > respect to ectutwfc. > > If we look at the 4 bands that are calculated we see the following > results: > > At ecutwfc=40 Ry: -5.1740 7.0370 7.0370 7.0370 > ecutwfc=100Ry: -5.1746 7.0369 7.0369 7.0369 > ecutwfc=200Ry:-5.1746 7.0369 7.0369 7.0369 > Now try >ecutwfc=2000Ry: -5.2862 6.9066 6.9066 10.2399 > > Now why do the values change? If I look at the output file for the > 2000 Ry case I see that there is a negative starting charge: > > Initial potential from superposition of free atoms > Check: negative starting charge= -0.057614 > > But still, why is this? The usual answer for a negative starting > charge is to increase the wavefunction cutoff, although I suspect > that's not the problem in this case. > > So besides the using of a 2000 Ry cutoff for silicon, what else is > wrong here? > > Best, > Brad > UC Berkeley > > > &control >prefix = 'si' >calculation = 'scf' >restart_mode = 'from_scratch' >wf_collect = .false. >tstress = .true. >tprnfor = .true. >outdir = './' >wfcdir = './' >pseudo_dir = './' > / > &system >ibrav = 0 >celldm(1) = 10.2612 >nat = 2 >ntyp = 1 >nbnd = 4 >ecutwfc = 2000.0 > / > &electrons >electron_maxstep = 100 >conv_thr = 1.0d-10 >mixing_beta = 0.7 >diago_full_acc = .true. > / > CELL_PARAMETERS cubic > 0.0 0.5 0.5 > 0.5 0.0 0.5 > 0.5 0.5 0.0 > ATOMIC_SPECIES > Si 28.086 Si.pz-vbc.UPF > ATOMIC_POSITIONS crystal > Si -0.12500 -0.12500 -0.12500 > Si 0.12500 0.12500 0.12500 > K_POINTS automatic > 1 1 1 0 0 0 > > > > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > -- - Prof Nicola Marzari Department of Materials Science and Engineering 13-5066 MIT 77 Massachusetts Avenue Cambridge MA 02139-4307 USA tel 617.4522758 fax 2586534 marzari at mit.edu http://quasiamore.mit.edu
[Pw_forum] What happens at REALLY large ectuwfc?
I happened to stumble upon something today that I thought was unusual. The problem is illustrated in a simple example: if you take the following basic cubic Si input file and, say, the Si.pz-vbc.UPF pseudopotential from the QE website and do a test for convergence with respect to ectutwfc. If we look at the 4 bands that are calculated we see the following results: At ecutwfc=40 Ry: -5.1740 7.0370 7.0370 7.0370 ecutwfc=100Ry: -5.1746 7.0369 7.0369 7.0369 ecutwfc=200Ry:-5.1746 7.0369 7.0369 7.0369 Now try ecutwfc=2000Ry: -5.2862 6.9066 6.9066 10.2399 Now why do the values change? If I look at the output file for the 2000 Ry case I see that there is a negative starting charge: Initial potential from superposition of free atoms Check: negative starting charge= -0.057614 But still, why is this? The usual answer for a negative starting charge is to increase the wavefunction cutoff, although I suspect that's not the problem in this case. So besides the using of a 2000 Ry cutoff for silicon, what else is wrong here? Best, Brad UC Berkeley &control prefix = 'si' calculation = 'scf' restart_mode = 'from_scratch' wf_collect = .false. tstress = .true. tprnfor = .true. outdir = './' wfcdir = './' pseudo_dir = './' / &system ibrav = 0 celldm(1) = 10.2612 nat = 2 ntyp = 1 nbnd = 4 ecutwfc = 2000.0 / &electrons electron_maxstep = 100 conv_thr = 1.0d-10 mixing_beta = 0.7 diago_full_acc = .true. / CELL_PARAMETERS cubic 0.0 0.5 0.5 0.5 0.0 0.5 0.5 0.5 0.0 ATOMIC_SPECIES Si 28.086 Si.pz-vbc.UPF ATOMIC_POSITIONS crystal Si -0.12500 -0.12500 -0.12500 Si 0.12500 0.12500 0.12500 K_POINTS automatic 1 1 1 0 0 0 -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20100122/728285a7/attachment.htm